Semi-continuous anaerobic digestion of a food industry wastewater inan anaerobic ®lter
S. Di Berardinoa, S. Costaa, A. Convertib,*
a Instituto Nacional de Engenharia e Tecnologia Industrial, Azinhaga dos Lameiros �a Estrada do Pacßo do Lumiar, 1699 Lisboa Codex, Portugalb Dipartimento di Ingegneria Chimica e di Processo ``G.B. Bonino'', Universit�a di Genova, Via Opera Pia, 15, I-16145 Genova, Italy
Received 14 January 1999; received in revised form 8 May 1999; accepted 14 May 1999
Abstract
The experimental results of semi-continuous tests of anaerobic digestion of confectionery industry wastewater, carried out at
di�erent residence times and organic loads in an up¯ow anaerobic ®lter, are presented and discussed. Giving COD removals higher
than 80% under the whole range of conditions tested, the anaerobic ®lter demonstrated not only a great ability of biomass to adapt
itself to a new carbon source but also an excellent capability to deal with organic load ¯uctuations and to utilise dilute feeds.
Sampling at di�erent ®lter heights demonstrated that the biogas development ensured mixing within the ®lter and that most of the
organic substances were utilised at the bottom of the reactor, especially when very dilute wastewater was fed. The results of this
work could be taken as a starting basis for scaling-up the process to the industrial scale. Ó 1999 Elsevier Science Ltd. All rights
reserved.
Keywords: Anaerobic hybrid ®lter; Anaerobic digestion; Food industry wastewater; Biogas production
1. Introduction
Anaerobic processes can favourably compete with thetraditional aerobic processes for the treatment of foodindustry wastewater, provided that the wastewater fromthe industrial activity is su�ciently concentrated, avail-able at relatively high temperature (Sixt and Sahm,1987), and characterised by high biodegradability(Mata-Alvarez et al., 1992). In those cases where thee�uents are quite dilute, with CODs from 2 up to 4times higher than those usually present in domesticwastewater, the aerobic processes can be too expensiveand then the anaerobic digestion can constitute an in-teresting alternative.
The present case study deals with the e�uent comingfrom a Portuguese confectionery factory, with averagepolluting power ranging from 0.63 to 2.62 g-COD/l. Theexisting completely-mixed digester is not able to biode-grade it e�ectively because the factory e�uent is quitedilute and the residence time too short (2.0 d). As aconsequence of this, the speci®c growth rate of biomassis lower than the dilution rate, thus preventing the de-velopment of a stable and abundant methanogenic
population. For this reason, the reactor needs to bemodi®ed so that the anaerobic biomass can adequatelybe retained independently of the wastewater ¯owrate.
Among the possible alternatives, one of the most at-tractive would be the adoption of the ¯uidised-bed re-actor (Hickey and Owens, 1981). Excellent results wereobtained both at bench- and pilot-scales for the anaer-obic digestion of dilute (Switzenbaum and Jewell, 1978)and concentrated (Converti et al., 1993) domestic sew-age, of e�uents from washing operations of industrialwine production (Converti et al., 1990), and of whey(Boening and Larsen, 1982). Methane yields alwaysexceeding 85% of the maximum theoretical value wereachieved in all those cases.
Since, however, industrial applications of this tech-nology appear to be too complicated, a much simplersolution is the anaerobic ®lter. In this reactor con®gu-ration, anaerobic bacteria grow abundantly attached toan inert support and form aggregates in the shape of¯ocs. These are retained and concentrated in the reactor,thus allowing a biomass retention time longer than thehydraulic residence time. The convective motion origi-nated by the biogas production ensures the agitation ofthe digesting mass. This type of reactor has already beenutilised with success for the anaerobic digestion of dif-ferent e�uents in the presence of di�erent supports
Bioresource Technology 71 (2000) 261±266
* Corresponding author.
0960-8524/00/$ ± see front matter Ó 1999 Elsevier Science Ltd. All rights reserved.
PII: S 0 9 6 0 - 8 5 2 4 ( 9 9 ) 0 0 0 8 0 - 2
(Bonastre and Paris, 1989; Marques et al., 1986±1988;Oliveira Marques et al., 1989; Solisio and Del Borghi,1987; Solisio et al., 1987).
In order to obtain the ®gures necessary to modify theexisting well-mixed reactor into an anaerobic ®lter, theexperimental results of anaerobic digestion in a bench-scale version of this equipment are presented and dis-cussed. To this purpose, this food industry e�uent hasbeen digested at di�erent loads and residence times.
2. Methods
2.1. Feed preparation
Several batches of e�uents from the industrial in-stallation were collected before entering the existingtreatment plant and transferred into storage vessels withvolume of 15 l. In order to have a constant compositionof the feed and to stop its spontaneous degradation, thebatches collected during di�erent 8 day-periods weremixed, stored in a room refrigerated at 4°C, and thenused as feed for the bench-scale equipment.
The average compositions of these feeds (Table 1),which were analysed in quadruplicate before startingeach trial at a given residence time, show that they wererelatively dilute, scarcely stabilised, mainly soluble, quitebiodegradable, and acidic, with the only exception beingthe last sample. Total solids and COD varied within theranges 1.10±2.30 g/l and 0.53±2.62 g/l, respectively, withthe lowest concentrations being detected in the last feedsdue to some reductions of activity in the factory.
2.2. Experimental set-up and digestion conditions
The up¯ow laboratory-scale column used for theanaerobic digestion tests was described in detail in aprevious paper (Di Berardino et al., 1997). The inertsupport was PVC tubes, which were cut to give 25±30mm-long pieces. In this way, the size of the ®lling ma-terial to be used in the industrial-scale plant (about 6 cm
in diameter) was simulated. The reactor, whose workingvolume was regulated at about 10 l, was surrounded bya water jacket to regulate the temperature at the oper-ating value (35°C). It was also provided with 4 holes forsampling at di�erent heights from the bottom (25, 50,75, and 100 cm).
The column, ®lled with two separate layers, each of0.2 m height, was fed semi-continuously by means of aperistaltic pump. The feed was left at room temperaturefor about 16 h and then neutralised with a solution of15% NaOH before use. Biogas production was followedby means of a wet gas meter provided with a totalizer.
2.3. Inoculum and start-up
To accelerate the start-up operation preceding thepseudo-steady-state conditions, a stable population ofbacterial ¯ocs from a bench-scale anaerobic ®lter wasused as inoculum. This anaerobic population had de-veloped within the bed during a period of two yearswhen the ®lter had previously been employed for thetreatment of milk-industry wastewater (Oliveira Mar-ques et al., 1989).
2.4. Analytical procedures
The course of digestion was followed by daily deter-mining in both the feed and the e�uent samples thecontents of total solids (TS), volatile solids (VS), totalsuspended solids (TSS), volatile suspended solids (VSS),chemical oxygen demand (COD), and volatile fatty acids(VFA), according to Standard Methods (APHA, 1985).
The methane content of the biogas at room temper-ature and pressure was daily determined by gas chro-matography, as explained in more detail in a previouswork (Di Berardino et al., 1997).
Following the experience of previous studies on an-aerobic digestion, waiting for at least ®ve days of rela-tively stable values of speci®c biogas production at agiven average residence time was the criterion used toget an indication of steady-state achievement.
Table 1
Average compositions of di�erent e�uents collected at various times from the real-scale system
No. of sample
1 2 3 4 5 6 7
pH 6.2 5.4 5.3 4.9 4.0 7.0
TS (g/l) 2.30 1.80 1.90 2.05 1.10 1.20
VS (g/l) 1.44 1.13 1.19 1.13 0.43 0.59
TSS (g/l) 0.65 0.38 0.19 0.26
VSS (g/l) 0.56 0.35 0.17 0.19
CODtot (g/l) 2.52 2.50 2.32 2.62 2.09 1.53 0.53
CODsol (g/l) 0.87 1.54
TS�Total solids; VS�Volatile solids; TSS�Total suspended solids; VSS�Volatile suspended solids; CODtot�Total COD concentration;
CODsol �Soluble COD concentration. Each sample, which was analysed in quadruplicate, consisted of a mixture of e�uents from the industrial
plant collected during di�erent 8 day-periods.
262 S. Di Berardino et al. / Bioresource Technology 71 (2000) 261±266
3. Results and discussion
3.1. Experimental schedule
The whole experimental programme was subdividedinto two separate phases. When the reactor was initiallyfed after a two-month stop, it rapidly began to workregularly. After only 5 days, it released a well puri®ede�uent and produced biogas in proportion to the or-ganic load. At the start of this ®rst phase, which lastedabout 35 days, the ®lter was semi-continuously fed at arelatively high average hydraulic residence time (5 days),corresponding to an average ¯ow-rate of 2.0 l/d. Pro-grammed variations of the pump capacity allowed the¯ow rate to vary between 0.85 and 3.85 l/d, thus simu-lating the organic load oscillations observed in the real-scale system (Fig. 1, curve (b)). During the second phaseof the experimental work, in order to study a wide rangeof di�erent conditions, the hydraulic residence time was
gradually lowered to 30 h, corresponding to a ¯ow rateof about 8.0 l/d.
In two di�erent periods of the experimental schedule,starting from days 74 and 113, respectively, the reactorfeeding was stopped for 10 and 6 days, in order tosimulate undesired occasional production stoppages ofthe real-scale plant.
3.2. Process e�ciency and e�uent characteristics
Table 2 summarises the conditions tested and themain results obtained under pseudo-steady-state condi-tions during the digestion of this food industry waste-water in the bench-scale anaerobic ®lter. Since thesamples were taken daily, the number of data used foreach mean value is coincident with the number of dayseach trial lasted. From these mean values and thoseshown in Fig. 2, it is evident that the e�ciency of or-ganic matter removal was very high at all the testedvalues of the organic load. In fact, not even when verydilute feeds were used did the e�ciency fall below 80%,while a maximum value higher than 92% was achieved.
The COD concentration of the treated e�uent neverexceeded 0.30 g/l, with the only exception being the testcarried out at the highest organic load. Besides thetypical performance worsening at high organic load, aninstantaneous discharge of organic matter during thissampling may have taken place. As shown in Fig. 3, thepH never decreased below 7.2, with a peak of 8.4, whichis indicative of a complete biodegradation of the organicacids, as it was con®rmed by periodic analyses of VFA.
3.3. Biogas production and characterisation
The methane content of the biogas and the speci®cbiogas production are plotted versus the organic load inFigs. 2 and 4, respectively.
Fig. 1. Variations of (a) the production of biogas and (b) the organic
load during the whole experimental programme.
Table 2
Trial conditions and steady-state results of the digestion of wastewater in a bench-scale anaerobic ®lter
Trial conditions
No. of sample 1 2 3a 3b 4 5 6 7
COD of the feed (g-COD/l) 2.52 2.50 2.32 2.32 2.62 2.09 1.53 0.53
Residence time (h) 133 83 51 53 51 44 35 31
Organic load (g-COD/l.d) 0.45 0.72 1.09 1.05 1.23 1.14 1.05 0.41
Trial duration (d) 14 24 12 6 18 a 22 22 b 10
Experimental results
E�uent COD (g-COD/l) 0.19 0.20 0.28 0.23 0.44 0.21 0.18 0.097
COD removal (%) 92.5 92.1 87.8 90.0 83.2 89.9 88.4 81.7
Biogas production (l/d) 0.99 1.34 3.56 4.00 3.85 3.98 2.66 1.84
Speci®c biogas production (l/g-
CODremoved)
0.24 0.20 0.37 0.42 0.38 0.39 0.29 0.55
Methane content of biogas (%) 85.9 87.3 86.1 86.1 84.3 86.5 89.9
Experimental results refer to mean data of samples daily collected along the entire duration of each trial at a given average residence time.a Not including 10 ®nal days of feeding stoppage.b Not including 6 days of feeding stoppage between days 114 and 119.
S. Di Berardino et al. / Bioresource Technology 71 (2000) 261±266 263
Biogas production ¯uctuated according to the or-ganic load (Fig. 1), thus showing conditions of greatstability for the system. The very low biogas productionobserved when the reactor feeding was stopped couldhave been the result of the ®nal degradation of the or-ganic substances previously introduced into the reactor.
The lowest values of the speci®c biogas productionwere obtained at the highest residence times (83 and 133h) corresponding to organic loads of 0.72 and 0.45g-COD/l.d, respectively. This result could have been dueto one or more of the following reasons: (a) accumula-tion of organic matter within the ®lter because of im-perfect mixing; (b) consumption of organic matter formaintenance instead of cell growth; (c) lack of processstability. At hydraulic residence times lower than 51 h,on the contrary, this process parameter achieved values(about 0.40 l/g-CODremoved) close to the maximum the-oretical methane production (0.35 l-CH4/g-CODremoved),thus suggesting a complete conversion of biodegradablesubstances into biogas.
Methane content of the biogas was very high (84.3±89.9%) under all conditions tested.
3.4. Activity distribution along the anaerobic ®lter
The microbial activity along the bed was studied bymeans of three series of TSS, VSS, and soluble CODdeterminations on samples withdrawn at 4 di�erent ®lterheights. Each series was characterised by a di�erentsoluble COD concentration in the feed (Table 3).
From these values on the whole it is evident that, atthe hydraulic residence times tested in this part of thestudy (30±51 h), most of the organic matter degrada-tion took place in the ®rst part of the ®lter bed (Fig. 5).In particular, when the most concentrated feed wasused, 85.5% of the soluble organic substances was re-duced within the ®rst 25 cm from the bottom. A sat-isfactory degree of reduction of the organic matter wasalso observed at the remaining three lateral sampling
Table 3
The concentrations of soluble COD and of suspended solids at dif-
ferent ®lter heights
Sample port Dimensionless
®lter height
TSS (g/l) VSS (g/l) CODsol (g/l)
Soluble COD of the feed: 1.59 g/l
4 1.00 0.125 0.105 0.125
3 0.75 0.136 0.122 0.129
2 0.50 0.278 0.240 0.181
1 0.25 0.150 0.140 0.232
Soluble COD of the feed: 1.33 g/l
4 1.00 0.075 0.075 0.098
3 0.75 0.065 0.065 0.088
2 0.50 0.132 0.128 0.122
1 0.25 0.115 0.110 0.146
Soluble COD of the feed: 0.53 g/l
4 1.00 0.015 0.015 0.057
3 0.75 0.015 0.015 0.052
2 0.50 0.060 0.050 0.057
1 0.25 2.145 1.615 0.057
TSS�Total suspended solids; VSS�Volatile suspended solids;
CODsol�Soluble COD concentration.
Fig. 2. In¯uence of the organic load on (n) the COD removal yield and
(D) the methane content of biogas.
Fig. 3. pH behaviour during the whole experimental programme.
Fig. 4. Speci®c biogas production (d) and COD concentration in the
e�uent (s) versus the organic load.
264 S. Di Berardino et al. / Bioresource Technology 71 (2000) 261±266
ports with the two most concentrated feeds. On theother hand, when using the less concentrated e�uent,the soluble COD kept at very low values (0.052±0.057g-COD/l) with negligible dependence on the ®lterheight, which means that the organic matter was nearlytotally degraded before reaching the ®rst samplingport.
The same type of organic matter distribution insidethe reactor, which strongly depended on the feed con-centration, has been observed for quite di�erent bio-logical processes, such as the treatment of gaseousstreams in a bio®lter (Zilli et al., 1996). In all cases, itwas possible to propose zero-order or ®rst-order de-gradation kinetics (Zilli et al., 1992, 1996). Mass velocitydistribution studies were also carried out, which sug-gested that the di�usion of substrate from the bulk tothe bio®lm could be the limiting step in most ®xed-bedprocesses (Converti et al., 1996, 1997).
The pro®les of TSS and VSS evidenced in Table 3demonstrate, for the two most concentrated feeds, theexistence of some agitation inside the ®lter, probablycreated by the remarkable biogas production whichre-suspended the solid matter near the second port. Afraction of these solids rose due to biogas develop-ment, whereas another fraction went down because ofthe degasi®cation provoked by the impact with thesupport. Anaerobic bacteria, as suggested by nearlycoincident values of VSS and TSS, mainly constitutedthis material. In the case of the most diluted feed, dueto its low organic content, the slow production ofbiogas was only able to expand the sludge bed exist-ing at the bottom. This expansion, which reached thelevel of the ®rst sampling port, was responsible for thehigh suspended solid concentration detected in thisregion.
Biogas production ¯uctuations (results not shown)were quite small compared with those detected for the
other parameters. Redox potential and pH assessmentsrevealed favourable average values for the process,namely ÿ268 mV and 6.94, while volatile acids kept atvalues always lower than the detection limit of the in-strumentation, thus suggesting their complete con-sumption at the bottom of the reactor.
References
APHA, 1985. Standard Methods for the Examination of Water and
Wastewater, 14th ed., APHA, AWWA, WPCF, Washington DC.
Boening, P.H., Larsen, V.F., 1982. Anaerobic ¯uidized bed whey
treatment. Biotechnol. Bioeng. 24, 2539±2556.
Bonastre, N., Paris, J.M., 1989. Survey of laboratory, pilot, and
industrial anaerobic ®lter installations. Proc. Biochem. 24, 15±20.
Converti, A., Zilli, M., Del Borghi, M., Ferraiolo, G., 1990. Fluidized
bed reactor in the anaerobic treatment of wine wastewater.
Bioproc. Eng. 5, 49±55.
Converti, A., Del Borghi, M., Ferraiolo, G., 1993. In¯uence of organic
loading rate on the anaerobic treatment of high strength semisyn-
thetic waste water in a biological ¯uidized bed. Chem. Eng. J. 52,
B21±B28.
Converti, A., Casagrande, M., De Giovanni, M., Rovatti, M., Del
Borghi, M., 1996. Evaluation of glucose di�usion coe�cient
through cell layers for the kinetic study of an immobilized cell
bioreactor. Chem. Eng. Sci. 51, 1023±1026.
Converti, A., Del Borghi, M., Zilli, M., 1997. Evaluation of phenol
di�usivity through Pseudomonas putida bio®lms: application to the
study of mass velocity distribution in a bio®lter. Bioproc. Eng. 16,
105±114.
Di Berardino, S., Bersi, R., Converti, A., Rovatti, M., 1997. Starting-
up an anaerobic hybrid ®lter for the fermentation of wastewater
from food industry. Bioproc. Eng. 16, 65±70.
Hickey, R.F., Owens, R.W., 1981. Methane generation from high-
strength industrial wastes with the anaerobic biological ¯uidized
bed. Biotechnol. Bioeng. Symp. 11, 399±413.
Marques, I.P., Di Berardino, S., Novais, J.M., 1986. Anaerobic
biodegradation in ®xed bed reactors. Int. Congr. on Renewable
Energy Resource, Madrid, 18±23 May 1986.
Marques, I.P., Caetano, A.T., Novais, J.M., 1987. A pig slurry
digestion study using up¯ow anaerobic ®lters. In: G. Grassi et al.
(Ed.), Biomass for Energy and Industry. Elsevier Applied Science,
Amsterdam, pp. 190±194.
Marques, I.P., Pereira, P., Novais, J.M., 1988. Pig slurry anaerobic
digestion in ®xed bed reactor. Proceedings 5th Int. Symp. on
Anaerobic Digestion, Bologna, pp. 401±404.
Mata-Alvarez, J., Llabr�es, P., Cecchi, F., 1992. Anaerobic digestion of
the Barcelona central food market organic wastes: Experimental
study. Bioresource Technol. 39, 39±48.
Oliveira Marques, F., Di Berardino, S., Nogueira, M.H., Novais, J.M.,
1989. Diluted milk factory wastewater pre-treatment using anaer-
obic ®lters. 5th Eur. Conf. on Biomass for Energy and Industry,
Lisbon, 9±13 October 1989.
Sixt, H., Sahm, H., 1987. Biomethanation. In: Sidwich, J.M., Holdom
R.S. (Eds.), Biotechnology of Waste Treatment and Exploitation.
Ellis Horwood, Chichester, UK, pp. 147±172.
Solisio, C., Del Borghi, M., 1987. The anaerobic digestion of surplus
activated sludge in a ®xed-®lm bioreactor. Wat. Res. 21, 192±304.
Solisio, C., Palazzi, E., Del Borghi, M., 1987. Performance of an
anaerobic ®xed-®lm bioreactor for treatment of a settled domestic
sewage. Proceedings 4th Eur. Congr. on Biotechnol., vol. 1,
Amsterdam, 14-19 June 1987, pp. 214±217.
Switzenbaum, M.S., Jewell, W.J., 1978. Anaerobic attached ®lm
expanded bed reactor treatment of diluted organics. Proceedings
Fig. 5. Distribution of soluble COD concentration along the ®lter
height. Soluble COD of the feed (g-CODsol/l): (d) 1.59; (D) 1.33; (h)
0.53.
S. Di Berardino et al. / Bioresource Technology 71 (2000) 261±266 265
51st Ann. Wat. Pollut. Control. Fed. Conf., Anaheim, CA, USA,
October, 1978.
Zilli, M., Fabiano, B., Ferraiolo, A., Converti, A., 1996. Macro-kinetic
investigation on phenol uptake from air by bio®ltration. In¯uence
of super®cial gas ¯ow rate and inlet pollutant concentration.
Biotechnol. Bioeng. 49, 391±398.
Zilli, M., Lodi, A., Rubino, L., Converti, A., Ferraiolo, G., 1992.
Biological degradation of phenol in a one-stage bio®lter using
di�erent Pseudomonas strains. Proceedings 9th World Clean Air
Congress, vol. 4, Paper IU-18C.09, Montr�eal, 30 August±04
September 1992.
266 S. Di Berardino et al. / Bioresource Technology 71 (2000) 261±266
